Psychoacoustic Loss Functions

These are the slides and example codes of our part (G. Schuller, R. Profeta) of our Workshop "Teaching AI to Hear Like We Do: Psychoacoustics in Machine Learning" at the October 2022 AES Convention in New York. The collection of slides from all participants can be found at: https://www.aes.org/technical/documentDownloads.cfm?docID=774

Introduction slides:

Our section on perceptual loss functions, based on explicit or implicit psycho-acoustic models:

Code examples for different perceptual evaluation functions:

Find the code for the psycho-acoustic pre- and post-filters in:

https://github.com/TUIlmenauAMS/Python-Audio-Coder

The Colab Jupyter notebook for the perceptual loss function, using the psycho-acoustic model, with comparisons and explanations:

The perceptual loss function only:

[] (https://colab.research.google.com/github/TUIlmenauAMS/PsychoacousticLoss/blob/main/onlyPsyacLoss.ipynb)

Asilomar Conference on Signals, Systems, and Computers 2024

We organised a special session on "Perceptual and Higher Level Loss and Distance Functions for Audio and Acoustics", see an overview here: https://cmsworkshops.com/Asilomar2024/view_session.php?SessionID=1126

Our special session talk slides are "Asilomar2024PsyacLossTalk.pdf" in this repository, "asilomar2024_report.pdf" is an overview of our special session.

All talks of the special session can be found in the subdirectory https://github.com/TUIlmenauAMS/PsychoacousticLoss/tree/main/Asilomar2024SpecialSessionTalks